{Minutes} TTWG meeting 2020-01-21

Thanks all for attending today's TTWG meeting. Minutes can be found in HTML format at https://www.w3.org/2021/01/21-tt-minutes.html


In text format:

   [1]W3C

      [1] https://www.w3.org/


                Timed Text Working Group Teleconference

21 January 2021

   [2]Previous meeting. [3]Agenda. [4]IRC log.

      [2] https://www.w3.org/2021/01/07-tt-minutes.html

      [3] https://github.com/w3c/ttwg/issues/172

      [4] https://www.w3.org/2021/01/21-tt-irc


Attendees

   Present
          Andreas, Atsushi, Cyril, Gary, Glenn, Nigel, Pierre

   Regrets
          -

   Chair
          Gary, Nigel

   Scribe
          cyril, nigel

Contents

    1. [5]This meeting
    2. [6]Permit implementations to use fontSize and fontFamily
       when computing line height w3c/ttml2#1215
    3. [7]TTML2 - Exiting CR
    4. [8]Placeholder for workplan for this year
    5. [9]Meeting close

Meeting minutes

  This meeting

   Nigel: I've added to the agenda two TTML2 topics, and a
   placeholder for the workplan for this year.
   … The TTML2 topics are issue 1215 about lineHeight and Exiting
   CR.
   … Any other business?

   group: [no other business]

  Permit implementations to use fontSize and fontFamily when computing
  line height w3c/ttml2#1215

   github: [10]https://github.com/w3c/ttml2/issues/1215


     [10] https://github.com/w3c/ttml2/issues/1215


   Glenn: I'm able to open a PR maybe by tomorrow as per Nigel's
   most recent proposal. I'm fine with that approach.

   Nigel: Okay, great, thank you.

   Glenn: I wanted to add one comment, that I didn't necessarily
   make it clear.
   … Based on my analysis, my conclusion was that an
   implementation that does _not_ take font*** into account now is
   non-compliant.
   … Your original description was that some implementations might
   not follow that semantic, and my conclusion was that
   … not doing so would make it non-compliant according to the
   letter of the current spec.
   … I would agree that it is not easy to pull out that thread of
   logic from the spec because it requires a fairly good
   … understanding of XSL-FO, which since we're talking about CSS
   might make one wonder.
   … The semantic model is fairly clear.
   … The way we did the derivation of attributes might make it
   seem like it's not semantically tied to the XSL-FO definitions.

   Nigel: Thanks for that.

   SUMMARY: @skynavga to prepare a pull request on TTML2

  TTML2 - Exiting CR

   nigel: my goal was to find out what minimally we needed to do
   … the IR contains a large number of tests
   … what wasn't clear to me was how it related to the exit
   criteria
   … [reading the exit criteria]
   … there is a document that lists all the changes
   … I went through all of the substantive changes
   … and tried to map them to the features
   … some don't and some do
   … from each PR to the feature or features
   … what sort of tests do we need: validation vs presentation
   … there are 22 changes made
   … that related to a feature
   … there are fewer features because some changes relate to
   multiple features
   … when we look at all of the tests that are in the current IR,
   some of them test things that are not changed in the 2nd ed
   … they are good tests, nothing wrong, but they are not needed
   for the exit criteria
   … my goal is to simplify
   … [working on an example in the spreadsheet ...]

   nigel: looks possible indicate if we could pass exit criteria
   for a change

   glenn: in that language you quoted about the definition of what
   needs to be tested, keep in mind that the word 'feature' can
   mean one of 2 things and it's not defined precisely.
   … It could mean a semantic feature or a designated feature with
   a #
   … second comment is that when I created the tests my model was
   for which PR required a test, I used whether we decided if it
   was substantive or not
   … lastly, you said a validating processor can flag the error, I
   think you meant 'must' flag the error
   … otherwise it's not a compliant validator

   nigel: on the last one, it's not clear what a validation
   processor needs to do if the requirement is for the
   presentation processor to ignore

   glenn: we don't say that a validating processor must do X, we
   don't do that in general
   … definition of valid vs invalid is defined further up in the
   document
   … even though that particular text does not use a must, it is
   still a must if you adhere to other semantics

   nigel: on the point on the 'substantive' issues, I also
   factored that
   … I also found anyway a significant number of tests that don't
   relate to 2nd edition changes
  … we could remove them from the IR

   glenn: I'm surprised of that

   cyril: me too

   nigel: I do plan on sharing the spreadsheet

   glenn: one example would be interesting to check your logic

   nigel: there are a bunch of ISD related tests
   … and I don't think there were any substantive changes to ISD

   glenn: it had to do with the fact that the definition of the
   size attribute was changed to xsd:nonNegativeDefinition to a
   new one

  nigel: there was a change like that
   … maybe I misunderstood that

   glenn: all the tests are changing that semantic change in TTML2
   2nd ed
   … I'm pretty sure all the changes can be traced back to a real
   change
   … if not, let me know, I can try to give a rationale

   nigel: anyway there are definitely tests which we are not able
   to demonstrate indep implementations

   pal: another column could be added as ttval
   … which is a constrained validator for TTML2
   … I'm not sure it's worth spending a lot of time on ttval by if
   a minor change can clean some rows, maybe we should do it

   nigel: the other observation is about the structure of the IR
   … the structure of the IR doesn't match the exit criteria
   … in several cases, you discover that the features are
   implemented by 2 implementations but the current report does
   not show that
   … it should be more traceable and easier to see the results
   … that would help us check if we pass the exit criteria

   glenn: I recall that earlier we required 2 validation
   implementations and 2 presentation implementations
   … you're suggesting changing that?

   glenn: yes
   … I won't argue with that but for the 1st edition we went
   beyond that

   nigel: I haven't checked the 1st edition exit criteria but in
   this case we need 2 implementations total
   … we currently have 2 reports: TTV and BBC but we could report
   ttval and imscjs

   glenn: although I haven't filled in the table, I have made
   progress on some of them (shear, opacity, ruby ...)
   … I think there is only or 2 (font selection strategy
   lineheight normal) that I did not touch
   … I don't have audio playback on TTPE

   nigel: for audio, we probably already have a pass already
   because TTV does validation and Adhere passes presentation
   … this is one example of one feature where we have 2
   implementations

   glenn: this model of IR was created by Cyril by taking the
   first IR and plugging in the relevant entries for the 2nd
   edition
   … we had a stronger requirement in the 1st edition
   … I don't have any objections to use a new model for the IR

   cyril: it would be good to know if the tests are relevant for
   IMSC

   nigel: I think I have done that pass already
   … so I should have that information, but that's a good idea

   nigel: I also tried to map the current tests in the IR to the
   changed features
   … I was doing it manually and realized it's already in the JSON
   files
   … so my next task is try to extract the 'since2e' tests

   cyril: the purpose would be to rewrite the IR with feature,
   tests, implementations

   nigel: if anybody has lines of code, please let me know
   otherwise I'll go ahead

   glenn: we are using the same manifest for different versions

   nigel: it's really useful

   glenn: nothing is using and I haven't validated it

   nigel: any other thoughts/questions/comments on exit CR for
   TTML2?

   nigel: I'm reassured from this discussion that we can do this
   if we focus on it

  Placeholder for workplan for this year

   nigel: last time the things that people wanted to work on was
   WebVTT and TTML2 to Rec
   … and beginning the journey of user customization

   nigel: does anybody has anything else?

   glenn: I was watching an movie about the creation of the Oxford
   dictionnary that took a while and hope we can be faster than
   that

   pal: thanks so much for your work helping us get TTML2 to Rec
   … on the customization thing, it would be great to get the
   world to agree on a framework to customization
   … I would caution us to do that on our own
   … going down one path that works well for one constituent
   … anything we can do to get people on the table is great

   nigel: as having worked on a fork of IMSC.js to add some
   customization in a particular work, I could be one constituent
   for who it works but I'm aware it might not work for others and
   don't want to impose a "BBC" solution

   pal: one thing we could do is: many of us are involved in many
   forums and we could share that
   … we could also draft something and share through liaisons

   nigel: that'd be a good first step

   pal: the privacy aspect is not something that many folks think
   about
   … so including in the problem statement would be useful

   nigel: not everybody share the same view of the possible impact
   … it needs thoughts and considerations
   … but I don't think it's a problem that concerns everybody

   nigel: for people who were not at the meeting last time, and if
   you have topics to add, I'm happy to add them, discuss here or
   have a private chat
   … we are chartered till the end of this year
   … I wonder if there is a view that having got to IMSC1.2 and
   TTML2 that this group has completed its work
   … have a feeling that some people think that
   … and it does not feel right
   … it'd be interesting to know the problem space that needs
   further work

   pal: we've reached an inflexion point in market adoption
   … and until we have broad adoption, I don't expect more spec
   work
   … people in this call are pioneers
   … we'll see much much broader adoption
   … and as new people come in we'll have more issues
   … I expect this year and maybe the next year to be quiet
   … but I'm not a big fan of saying we're done

   gkatsev: I wonder if the work, the meta work TTML to WebVTT
   draft, could be interesting

   nigel: in terms of spec work, I'd like to push the Audio
   Description Profile of TTML2
   … I expect spec work
   … it falls in the category of additional work to use the core
   specs

   <atsushi> [11]https://www.w3.org/blog/2021/01/wcag-3-fpwd/


     [11] https://www.w3.org/blog/2021/01/wcag-3-fpwd/


   <nigel> Gary: I just saw the notice that WCAG 3 work has begun
   - maybe there's work we should do to help them with
   accessibility for subtitles and captions.

   nigel: they've just released the FPWD
   … it may have an impact on the customization work

   Cyril: In general the topic of subtitle quality is of interest
   to Netflix.
   … I don't know if the group would be happy to work in this
   direction,
   … facilitating verification of reading speed, or detections of
   errors, for example typos, missing words.
   … We had an interesting session at Netflix where we brought in
   Deaf customers and asked them what they were missing
   … when it comes to subtitles. Maybe we could do the same in
   W3C, ask users of subtitles what is missing,
   … what would help them, identify gaps that we could fill to
   help them.

   Nigel: That's an interesting point - our core constituency in
   W3C is the users and we have very little direct engagement from
   … users, so there's definitely something there.

   Cyril: W3C has access to a lot of communities, so maybe it is a
   good opportunity to reach out and find out if there are
   frustrations
   … with how subtitles are done in general.

   Andreas: That's a really important point - the question is if
   it is a topic for a technically focused group.

   Nigel: True, there may be a crossover for example - is there a
   technical solution to varying the amount of editing and the
   word rate?

   Atsushi: We could make a video to show at the AC meeting to
   show the possible directions of our work.

  Meeting close

   Nigel: We're a few minutes over today, so let's adjourn. Thanks
   everyone. [adjourns meeting]


    Minutes manually created (not a transcript), formatted by
    [12]scribe.perl version 127 (Wed Dec 30 17:39:58 2020 UTC).

     [12] https://w3c.github.io/scribe2/scribedoc.html